Anonymous Networking amidst Eavesdroppers

TL;DR

This paper analyzes security against timing-based traffic analysis in wireless networks, proposing an information-theoretic measure of anonymity and designing scheduling techniques to optimize network throughput while maintaining specified anonymity levels.

Contribution

It introduces an analytical measure of anonymity using equivocation and develops scheduling strategies that balance network performance with anonymity constraints.

Findings

Achievable rate regions are characterized for two-hop networks with maximum anonymity.

A scheduling strategy is designed to maximize throughput given an anonymity requirement.

The throughput-anonymity relationship is shown to be equivalent to a rate-distortion function.

Abstract

The problem of security against timing based traffic analysis in wireless networks is considered in this work. An analytical measure of anonymity in eavesdropped networks is proposed using the information theoretic concept of equivocation. For a physical layer with orthogonal transmitter directed signaling, scheduling and relaying techniques are designed to maximize achievable network performance for any given level of anonymity. The network performance is measured by the achievable relay rates from the sources to destinations under latency and medium access constraints. In particular, analytical results are presented for two scenarios: For a two-hop network with maximum anonymity, achievable rate regions for a general m x 1 relay are characterized when nodes generate independent Poisson transmission schedules. The rate regions are presented for both strict and average delay constraints on traffic flow through the relay. For a multihop network with an arbitrary anonymity requirement, the problem of maximizing the sum-rate of flows (network throughput) is considered. A selective independent scheduling strategy is designed for this purpose, and using the analytical results for the two-hop network, the achievable throughput is characterized as a function of the anonymity level. The throughput-anonymity relation for the proposed strategy is shown to be equivalent to an information theoretic rate-distortion function.